microRNA Biomarkers as Potential Diagnostic Markers for Cancer - Genomic Biomarkers for Pharmaceutical Development - page 94

5.2 D EVELOPMENT OF A miRNA-BASED CLINICAL T EST
5.2.1 Biomarker Discovery on a High Content Platform
Biomarker discovery plays a key role in the larger context of implementing a clinical test.
Since the goal is to implement a validated test, the platform to be used for the test is often differ-
ent from the discovery platform, and that sets the stage for a series of studies that have concep-
tually similar design qualities: sample size estimation; expression profiling; and performance
estimation ( Fig. 5.1 ). Sample size estimates for biomarker discovery using array technology
have been extensively evaluated in the literature [15-19] . However, to our knowledge, research
has not focused on incorporating the complexities and noise of assay into model migration.
Migrating a predictive model from a microarray (or next-generation sequencing) platform to
RT-qPCR, for example, is a challenge for clinical testing as novel design and analysis strategies
for performance preservation face increased scrutiny.
A high-throughput and large-content platform is used to generate a hypothesis, typically in
an unbiased manner where all possible genes (or miRNAs) are considered as possible candi-
dates for discriminating the groups of interest. Biomarker discovery and predictive modeling
has been rigorously applied to mRNA expression data derived from array technology [20-27] .
Many of the qualities of sound design and analysis strategies for successful mRNA biomarker
discovery also hold true for miRNA studies including:
● involvement of bioinformatics and biostatistics teams in the study design;
● balancing of classes with other relevant clinical covariates, particularly batching of
experiments and site of sample origin;
FIGURE 5.1 A high level overview of biomarker discovery in the context of platform migration.The platform
used for unbiased discovery and predictive modeling is typically different than the platform used for a clinical test.
As a result, a bridging study (or platform migration) occurs where results between the two platforms are shown to
be equivalent. For mRNA- and miRNA-based signatures and predictive models, the destination platform is typi-
cally one based on RT-qPCR. The figure does not capture the clinical and analytical validation of the final RT-qPCR
based test.